How to Rank Shopify Products by Conversion Rate (and Why Best Sellers Mislead)

Rank Shopify products by conversion rate, not just sales. How views, add-to-carts and orders fit together, the small-sample traps that fool merchants, and how blended scores fix them.

Most Shopify stores rank their collections by sales, and many of them are leaving money on the table because of it. When you rank Shopify products by conversion rate instead of, or alongside, raw sales, a different picture of your catalogue appears. Some of your best sellers turn out to be mediocre products that happen to get a lot of traffic, and some of your quiet products turn out to convert brilliantly when anyone actually sees them. This guide explains how conversion, views and add-to-carts fit together, the statistical traps that catch almost everyone who tries this, and how a blended score avoids them.

The product with the most sales isn’t always your best product

Sales are an outcome of two things multiplied together: how many people saw the product, and how many of those people bought it. A product can top the sales chart because it’s excellent, or simply because it’s seen more than anything else.

And on a Shopify store, being seen more than anything else is often an accident of position. The products in the first row of a collection get far more views than the products on the third screen, because most shoppers never scroll that far. If the first row was filled by a sales-based sort, then being in the first row produces more views, which produce more sales, which keep the product in the first row. It’s a loop, and it can hold a product at the top long after something better has joined the range.

Here’s an illustrative example. These numbers are invented to show the arithmetic, not taken from any real store.

• Product A: 3,000 views in 30 days, 45 orders. Conversion 1.5%. It sits in position two.
• Product B: 400 views in 30 days, 18 orders. Conversion 4.5%. It sits in position twenty-three.

Sorted by orders, A wins easily, 45 to 18. But B converts three times as well. If B got A’s 3,000 views and kept anything like its conversion rate, it would produce far more orders than A does. A sales sort will never find that out, because it will never give B the views. That’s the core argument for conversion-aware ranking: it finds the products that deserve more exposure, not just the ones that already have it.

What product conversion rate actually measures

Store-wide conversion rate, the figure on your Shopify dashboard, is sessions that ended in an order divided by all sessions. Product conversion rate is different and much more useful for merchandising. In SellRank, product conversion is orders that included the product divided by the product’s page views, over a 30-day window.

• Views come from SellRank’s own pixel, installed through Shopify’s Web Pixels API, which counts product page views per product per day.
• Orders come from your Shopify orders, so they match your admin exactly.
• The window is 30 days, which is long enough to smooth out day-of-week swings and short enough to reflect the current range, prices and photography.

It measures how persuasive a product is once a shopper is looking at it: the photography, the price, the description, the reviews, the sizes available, the delivery promise. That’s exactly the quality you want to know about when you decide what deserves the first row.

The funnel inside every product: views, add-to-carts, orders

Conversion is the end of a small funnel, and each stage tells you something different.

• Views: does the product earn the click from the collection page? This is mostly about the tile: the image, the title and the price shown on the grid.
• Add-to-carts: once someone is on the product page, do they want it? This is about the page itself: the photos, the details, the sizing, the reviews.
• Orders: do they follow through? This is about everything after the cart: shipping costs, delivery time, payment options, second thoughts.

Looking at the stages separately helps you diagnose. A product with lots of views and few add-to-carts has an attractive tile and a disappointing page, or a price that looks different on the page than the shopper expected. A product with healthy add-to-carts and few orders may have a stock or sizing problem, or a shipping cost that appears late. A product with few views and a strong conversion rate is the hidden gem this article is about.

SellRank exposes each stage. Views and add-to-carts are available as strategies over 7 or 30 days. Conversion % is orders over views for 30 days. And cart-to-view %, add-to-carts divided by views, is available as a condition, so you can require a slot’s products to clear a level of intent before they qualify.

Four product patterns, and what to do about each

Once you can see views, add-to-carts and orders side by side for every product in a collection, most products fall into one of four patterns. Each calls for a different merchandising response, and only one of them is fixed by moving the product up the page.

• High views, high conversion: your genuine stars. They deserve the first row, and a sales sort usually already puts them there. Keep them in stock and protect their position.
• High views, low conversion: the tile promises more than the page delivers. Moving the product higher will only waste more attention. Look at the product page instead: photos, price, sizing, reviews, delivery promise. Until it improves, let it sit lower.
• Low views, high conversion: the hidden gems. The product persuades the few people who find it. This is the pattern conversion-aware ranking exists to fix: give it exposure and see whether the rate holds.
• Low views, low conversion: the tail. Neither the tile nor the page is working. Consider new photography, a price review, or a quieter position while you decide whether it belongs in the range.

A sales-only sort sees none of this. Orders are views multiplied by conversion, so a sales ranking can’t tell a genuine star from a high-traffic product that converts poorly but gets enough clicks to pile up orders anyway. It never notices the hidden gems at all, because they don’t have the views to accumulate sales. That’s why collection pages end up with persuasive products buried under ones that attract clicks and then disappoint. Separating views from conversion is the only way to tell these patterns apart, and it takes product-level view counts that Shopify’s collection sort simply doesn’t use.

Why add-to-carts are the underrated signal

Orders are the truth, but they’re slow. Many stores only get a handful of orders per product per month, which makes order-based rankings jumpy and slow to notice change. Add-to-carts happen more often than orders, because not every cart becomes an order, which means they accumulate into usable data much sooner.

That makes add-to-carts a leading indicator. When a product starts getting added to carts more often, orders usually follow. Ranking on 7-day add-to-carts, or blending them into a score, lets a collection respond to rising demand before it shows up in the sales column. SellRank’s Trending preset leans on exactly this: 7-day views 45, 7-day add-to-carts 35 and 30-day orders 20.

Trap 1: the tiny sample

Here’s where most attempts at conversion ranking fall over. Conversion is a ratio, and ratios on small numbers are wildly unstable.

• A product with 2 views and 1 order has a 50% conversion rate.
• A product with 1,000 views and 40 orders has a 4% conversion rate.

Sort naively by conversion and the first product goes to the top of the page. But nobody should believe that 50%. One more view without an order and it drops to 33%. Two more and it’s 25%. The second product’s 4% is based on enough evidence to take seriously. The first product’s 50% is based on one person.

SellRank guards against this with a simple floor. Below five views in the 30-day window, conversion and cart-to-view read zero, so a product with one lucky order on two views can’t leap to the top. Five views is a deliberately low bar. It removes the most absurd cases without excluding new or niche products for weeks. The floor is a guard, not a solution, and it’s why conversion works best as part of a blend rather than on its own.

Trap 2: regression to the mean

Even above the floor, extreme conversion rates on modest samples tend to drift back towards normal as more data arrives. A product with 20 views and 3 orders shows 15%. Perhaps it really is that good. More often, it had a lucky week, and when it gets 200 views its conversion will settle somewhere much closer to your store’s typical range.

The practical consequence: if you promote products to the top purely because their conversion is high, you’ll often be disappointed when they get there. The extra views will reveal a more ordinary rate. That doesn’t mean conversion is useless. It means you need something that weights conversion by the amount of evidence behind it.

Trap 3: position bias

Where a product sits on the page changes who sees it, and that changes its conversion rate. Products in the first row are seen by everyone, including casual browsers with low intent, so their conversion can look lower than their true appeal. Products deep in the page are mostly seen by determined shoppers who scrolled or searched to find them, so their conversion can look higher.

This cuts both ways. It means a deep-in-the-page product with strong conversion may not keep that rate at the top. It also means a top-row product with modest conversion may be doing better than it looks. The fix isn’t to ignore conversion but to compare like with like where you can, and to watch what happens after a change. That’s what per-version performance is for, which we come to below.

Trap 4: traffic mix

Products don’t all get the same kind of visitor. A product featured in a paid social campaign might get thousands of views from people who’ve never heard of your brand, and convert modestly. A product found through organic search for its exact name converts much better, because the visitor already wanted it. Neither rate is wrong, but they aren’t measuring the same thing.

If a product’s conversion suddenly drops, check whether its traffic suddenly grew from a new source before concluding the product has got worse. A 30-day window helps here, because it smooths short campaign bursts into a longer average.

Trap 5: price and basket effects

Cheaper products usually convert better than expensive ones, simply because the decision is smaller. If you rank a mixed-price collection purely by conversion, the top can fill with low-priced products, which may not be what the business wants. Revenue per view, or conversion balanced against revenue, gives a fairer picture across price points.

The same applies to products that are often bought together or as add-ons. A small accessory might convert brilliantly because people add it to a larger order. That’s valuable, but it’s a different role from a hero product, and it probably belongs in a different slot.

The fix: blended scores

Every trap above has the same root cause: a single ratio carries too much weight. The answer is to combine conversion with the volume signals that tell you how much to trust it. That’s what SellRank’s Balanced score does.

A Balanced score mixes several metrics into one score from 0 to 100, each with a weight you choose. For conversion-aware ranking, a sensible starting blend might be conversion 40, 30-day add-to-carts 30 and 30-day views 30. A product needs to convert well and have enough activity to be trusted to score highly. A product with a spectacular rate on thin data scores well on one ingredient and poorly on two, and lands in the middle, which is where it belongs until it proves itself.

Two design choices make SellRank’s blend behave well:

• Rank normalisation. Each product’s contribution from a metric depends on where it ranks among the products being sorted, not on how far ahead the leader is. With min-max scaling, one viral product with ten times the views of anything else would squash everyone else’s view score towards zero. With rank normalisation, it simply ranks first on views, and the rest spread evenly behind it.
• Shared ties. Products with identical values, including the many with no data, share their average rank. They stay level with each other instead of being scattered arbitrarily.

The result is a ranking where being good at several things beats being extreme at one, which is how an experienced merchandiser would judge a product anyway.

A worked blend, step by step

Here’s how a rank-normalised blend treats five products. The numbers are illustrative, invented to show the mechanics. The blend is conversion 40, 30-day add-to-carts 30, 30-day views 30.

• Product P: 2,400 views, 150 add-to-carts, 36 orders. Conversion 1.5%.
• Product Q: 900 views, 170 add-to-carts, 32 orders. Conversion about 3.6%.
• Product R: 350 views, 40 add-to-carts, 14 orders. Conversion 4%.
• Product S: 60 views, 9 add-to-carts, 4 orders. Conversion about 6.7%.
• Product T: 4 views, 1 add-to-cart, 1 order. Conversion reads zero, because it’s below the five-view floor.

Rank each metric. On conversion the order is S, R, Q, P, T. On add-to-carts it’s Q, P, R, S, T. On views it’s P, Q, R, S, T. With rank normalisation, first place on a metric is worth the full weight, last place is worth nothing, and the others are spaced evenly between.

• P: fourth on conversion, second on carts, first on views. Strong volume, weak persuasion.
• Q: third on conversion, first on carts, second on views. Good at everything.
• R: second on conversion, middle on carts and views. Persuasive with moderate evidence.
• S: first on conversion, fourth on carts and views. Very persuasive on thin evidence.
• T: last on everything. No usable evidence yet.

Work the weights through and the scores come out Q 72.5, P 62.5, R 60, S 55 and T 0. That’s a sensible outcome. Q is the product you’d bet on: it converts more than twice as well as P, draws more add-to-carts, and has plenty of activity to back it up. P still ranks second because its volume is real. S, the 6.7% product on 60 views, sits in the middle rather than at the top, where a pure conversion sort would have put it. If S keeps converting as its views grow, its volume ranks will rise and it will climb. T waits for data. Sorted by orders alone, P would lead on 36 orders to Q’s 32, and Q’s much better conversion would never have counted for anything. Sorted by conversion alone, S would lead on 60 views. The blend avoids both mistakes.

Choosing the window: 7 days or 30 days

SellRank calculates conversion over 30 days, and offers views, add-to-carts, orders and revenue over both 7 and 30 days. The choice of window for the volume ingredients matters as much as the weights.

• Seven-day windows react quickly. They’re right for high-traffic collections, fast fashion drops, and anything where this week genuinely differs from last month. On thin data they’re noisy.
• Thirty-day windows are steadier. They’re right for most collections on most stores, and essential for high-price, low-volume catalogues where a week might contain one or two orders per product.
• Mixing windows is often best. The Trending preset uses 7-day views and add-to-carts, which build quickly, with 30-day orders, which need longer to be meaningful.

Variants, sizes and conversion

For products with sizes or other variants, conversion is partly a stock story. A dress that’s sold out in the two most common sizes will convert poorly, not because shoppers dislike it but because most of them can’t buy it. Its conversion rate falls, its ranking slips, and that’s arguably the right outcome while the sizes are missing: a product most people can’t buy shouldn’t lead the page.

The thing to watch for is misreading that drop as a verdict on the product. When sizes are restocked, conversion should recover, and a 30-day window will take some time to reflect it. If you know a restock has landed, a temporary pin or placement can give the product its visibility back while the numbers catch up.

When conversion ranking is the wrong tool

Conversion isn’t the right lens for every collection. A few cases where it misleads:

• New Arrivals: new products have little data by definition. Rank by newness blended with early views and add-to-carts, as the Fresh + popular preset does.
• Clearance: the goal is moving stock, not finding persuasive products. Discount, stock cover and views matter more, as in the Clearance push preset.
• Very small collections: with a dozen products and modest traffic, there’s rarely enough data for conversion to separate them reliably. A simple revenue or orders ranking with a pin or two is often enough.
• Collections dominated by one traffic source: if most views come from one campaign, conversion reflects the campaign’s audience more than the products.

Presets that already include conversion

You don’t have to design a conversion blend from scratch. Two of SellRank’s presets already include it:

• Customer favorites: review rating 35, 30-day orders 35, conversion 30. Products people buy, like and convert on. A strong choice for beauty, gifting and anywhere social proof matters.
• Profit maximizer: 30-day profit 45, conversion 30, 30-day views 25. Products that make money and persuade shoppers, with enough traffic to trust. Needs cost per item and is included in Scale and Max.

Every preset is editable. Change a weight and the live preview reorders immediately, with each product’s score on its badge, so you can see exactly who moved and why before anything goes live.

Using conversion in percentage slots

Conversion is often most useful not as the whole page’s order but as the logic of one zone. SellRank’s percentage slots let you do that. Here’s an illustrative composition for a 100-product collection:

• Slot 1, 15%, Proven converters: conditions inventory units greater than 0 and 30-day views at least 50; ranking conversion %. The views condition means only products with a meaningful sample qualify, which deals with the small-sample trap directly.
• Slot 2, 10%, Hidden gems: conditions available to buy, cart-to-view % at least a level you choose; ranking Balanced score of conversion 50 and 30-day add-to-carts 50. Products that persuade the people who find them, given a chance at more exposure.
• Everything else: Revenue driver (30-day revenue 40, 30-day orders 35, 30-day views 25).
• Out-of-stock: parked at the end.

The views threshold of 50 is an example, not a recommendation. Choose one that suits your traffic. A high-traffic store might use several hundred; a small store might use 20. The principle is to require enough evidence before a ratio decides the top of the page.

Giving hidden gems a fair trial

The most valuable thing conversion ranking does is surface products that deserve more exposure. The second most valuable thing is to find out quickly when they don’t. A good process looks like this:

• Identify candidates: products with strong conversion or cart-to-view rates but modest views. The hidden gems slot above does this automatically.
• Promote them into a visible zone for a fixed period, such as two weeks.
• Watch what happens to their conversion as views rise. If it holds up reasonably well, they’ve earned their place. If it falls back to ordinary, the original rate was luck or position bias, and the blend will naturally move them down as the numbers update.

Because the Balanced score recalculates from live data, this happens without anyone manually shuffling products. Products that justify their promotion keep scoring well. Products that don’t, drift down.

Checking product conversion by hand first

If you’d like to see the effect before installing anything, you can do a rough version by hand. Take your busiest collection and list its top 20 products by orders over the last 30 days from Shopify Analytics. Then find product page views for the same period from whichever analytics you use, divide orders by views for each product, and sort the list by the result. Mark any product with very few views, because its rate can’t be trusted yet.

Merchants who do this often find at least one product that converts noticeably better than products sitting above it on the page. The exercise takes an hour and goes out of date the moment you finish it, which is the argument for letting software keep it current. But it’s a useful way to check whether the idea applies to your catalogue before you commit to anything.

Proving that conversion-aware ranking helped

Switching a collection from best selling to a conversion-aware blend is a hypothesis: that giving exposure to more persuasive products will turn the same traffic into more orders. It should be tested, not assumed.

SellRank records every published configuration as a version and attributes the collection’s daily views, add-to-carts, orders and revenue to whichever version was live. After a week or two on the new blend, the performance view shows it side by side with the previous version. If orders per view rose, you have evidence. If they didn’t, you can restore the old version in one click and try a different blend. It isn’t a randomised experiment, and seasonality still matters, but it’s a record you can reason from rather than a feeling. Performance analytics and version history are included from the Grow plan. On the free plan you can still run a simple before-and-after by comparing the collection’s numbers in Shopify Analytics for the week before and the week after your sync.

Why this needs first-party data

Product conversion depends on accurate view counts. If views are undercounted, conversion looks inflated; if they’re counted inconsistently across products, the ranking is distorted. Tools that rank by engagement often read views from Google Analytics, which is affected by consent choices, blocked scripts and configuration drift.

SellRank counts views itself with a first-party pixel installed through Shopify’s Web Pixels API. It counts product views and add-to-carts per product per day, as aggregates with no customer identifiers, and it’s consistent across every product in your catalogue. Orders come from Shopify. On Scale and Max, a GA4 property can backfill up to 90 days of history so conversion ranking works from day one, but the pixel remains the source of truth.

Conversion ranking by store type

• High-traffic fashion: plenty of views, so conversion stabilises quickly. A proven converters slot with a high views threshold works well at the top.
• Low-traffic, high-price jewellery or furniture: few orders per product. Lean on add-to-carts and cart-to-view, use 30-day windows, and keep conversion as a smaller ingredient in a blend.
• Beauty and wellness: reviews matter, so Customer favorites is a natural starting point.
• Stores with heavy paid social traffic: expect conversion to vary with campaign timing. Use 30-day windows and compare versions over full campaign cycles.
• Mixed-price catalogues: balance conversion with revenue so cheap products don’t take over the first row.

A checklist to rank Shopify products by conversion rate

• Measure views consistently for every product, ideally with a first-party pixel.
• Use a 30-day window for conversion unless your traffic is very high.
• Never rank on conversion alone without a minimum sample. SellRank’s five-view floor is a starting guard; a views condition on the slot is a stronger one.
• Blend conversion with views and add-to-carts so rates are weighted by evidence.
• Watch for position bias, traffic mix and price effects before drawing conclusions.
• Give promoted hidden gems a fixed trial period and let the blend move them if they don’t hold up.
• Keep the first row in stock, whatever the ranking.
• Record each version and compare them before deciding what to keep.

Why SellRank is built for conversion-aware merchandising

Ranking by conversion properly needs several things at once: consistent first-party view data, orders straight from Shopify, a sample floor, a blend that weights ratios by evidence, slots that can apply conversion to one zone of the page, a human override for what the data can’t know, and a record of what each version did. SellRank has each of those. Its 19 strategies include views, add-to-carts and orders over two windows and conversion over 30 days; its Balanced score is rank-normalised with presets that include conversion; its slots accept view and cart-to-view conditions; and every published order is verified in Shopify and tracked as a version.

Find your hidden best sellers

Somewhere in your busiest collection there’s probably a product that converts better than the ones above it and just isn’t getting seen. The quickest way to find it is to open SellRank on the free plan, create a smart sort on that collection with a conversion-aware blend, and read the preview. You’ll see which products the blend would promote and why, before anything changes on your store. Sync once to try it live. Plans and the install link are on the SellRank product page.